关于「林格曼效应与布鲁克斯定律:人越多,每人越少?」,《规模法则与组织新陈代谢》给出的定量结论是什么?

结论:林格曼的表显示人均出力随人数线性下降,等价于每条沟通通道固定损失,最优人数约为 7;但英厄姆等人发现损失在三人后趋于稳定,而且相当一部分来自动机而非协调。布鲁克斯定律是有条件的经验法则:项目末期、上手慢、沟通成本高时,加人会推迟完工;早期加人通常有利。下一章把视野从一个团队扩大到整个层级组织。

数据口径:本书讨论生物异速生长与城市、企业规模的数学模型,不是投资建议,也不构成任何企业管理成效的承诺。 书中以标注框区分已发表的证据、可运行的 Python 计算(15 段程序,全部核对)、原典与作者的假说。

English version

对照数据表

下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库

fitted loss per extra worker c = 0.0737; beta = 2c = 0.1474; N* = 1/beta + 1/2 = 7.28
Nper worker (table)total (table)total (fit)channels N(N-1)/2plateau model
11.001.001.0001.00
20.931.861.8511.75
30.852.552.5632.48
40.773.083.1263.24
50.703.503.53104.02
60.633.783.79154.81
70.563.923.90215.60
80.493.923.87286.40
12 people: all-to-all channels 66 | three teams of four plus three lead-to-lead links 21
Graicunas (1937) relationships for a manager with 4 direct reports: 44; with 12: 24708

核心公式

\[Y(N) = \alpha N - \frac{\beta}{2} N (N - 1), \qquad \frac{\mathrm{d}Y}{\mathrm{d}N} = \alpha - \beta\left(N - \tfrac12\right) = 0 \;\Rightarrow\; N^* = \frac{\alpha}{\beta} + \frac12 .\]
查看完整验算程序
# Ringelmann's (1913) summary table, translated by Kravitz & Martin 1986: relative output
# per worker for groups of 1 to 8 (often cited as rope pulling; Ringelmann did not name
# the task). Fit per-capita output y = 1 - c (N - 1); total output
# Y = N y = alpha N - (beta/2) N (N - 1) with alpha = 1, beta = 2c, i.e. a fixed loss per
# communication channel; peak at N* = alpha/beta + 1/2. Then: what if the per-capita loss
# levels off after three people (the curvilinear shape reported by Ingham et al. 1974;
# the plateau value 0.80 below is ILLUSTRATIVE)? Finally: channel counts for 12 people.
import math
PER_WORKER = [1.00, 0.93, 0.85, 0.77, 0.70, 0.63, 0.56, 0.49]

if __name__ == "__main__":
    ns = range(1, 9)
    c = (sum((n - 1) * (1 - y) for n, y in zip(ns, PER_WORKER))
         / sum((n - 1) ** 2 for n in ns))             # least squares through (1, 1)
    beta = 2 * c
    print(f"fitted loss per extra worker c = {c:.4f}; beta = 2c = {beta:.4f};"
          f" N* = 1/beta + 1/2 = {1 / beta + 0.5:.2f}")
    print("N   per worker (table)  total (table)  total (fit)  channels N(N-1)/2"
          "  plateau model")
    for n, y in zip(ns, PER_WORKER):
        plateau = n * (1 - 0.20 * (1 - math.exp(-(n - 1) / 1.0)))
        print(f"{n:2d}        {y:.2f}             {n * y:4.2f}          "
              f"{n - beta / 2 * n * (n - 1):4.2f}            {n * (n - 1) // 2:3d}"
              f"            {plateau:5.2f}")
    print("12 people: all-to-all channels", 12 * 11 // 2,
          "| three teams of four plus three lead-to-lead links", 3 * 6 + 3)
    k = 4
    print(f"Graicunas (1937) relationships for a manager with {k} direct reports:"
          f" {k * (2 ** (k - 1) + k - 1)}; with 12: {12 * (2 ** 11 + 11)}")

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《规模法则、异速生长与组织新陈代谢的统计物理》第 5 章「林格曼效应与布鲁克斯定律:人越多,每人越少?」(版本 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.23788553

DOI: 10.5281/zenodo.23788553 · 阅读本章所在专著 · 5DL-B-036 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026036,
  author    = {Fan, Yafu},
  title     = {Scaling Laws, Allometric Growth and the Statistical Physics of Organisational Metabolism},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  doi       = {10.5281/zenodo.23788553},
  url       = {https://doi.org/10.5281/zenodo.23788553}
}

本书其他章节结论